Van der Waals crystal mimics human neurons, advancing AI hardware with light-based learning

Breakthrough in Artificial Neuronal Cell Mimicking Using Light

A research team led by Professor Taesung Kim at Sungkyunkwan University has developed an optoelectronic synaptic device that replicates human neuronal and synaptic functions at the device scale using optical stimuli1

1

. The team designed a designable van der Waals crystal through a single-step sulfurization process involving mixed plasma, offering a structural solution for configuring semiconductor materials for brain-inspired computing2

2

. This development addresses critical challenges in creating next-generation neuromorphic semiconductors capable of processing vast amounts of visual data in real time.

Single-Step Process Transforms Material Structure

The researchers applied an argon and hydrogen sulfide plasma sulfurization process to bulk van der Waals rhenium selenide (ReSe₂), transforming the upper portion into a nanocrystalline ReSe₂ layer composed of nano-sized grains while preserving the underlying bulk single-crystalline ReSe₂ layer1

1

. These two integrated layers structurally correspond to the light-sensitive ion channels of a neuronal cell membrane and the intracellular environment, respectively, and were fabricated without additional deposition or patterning steps2

2

. This approach overcomes technical challenges including difficulty in grain boundary control, polymer residue accumulation, mechanical warpage at interfaces, and poor crystalline uniformity that plagued conventional van der Waals materials.

Enhanced Performance Through Controlled Ionic Migration

Using scanning probe microscopy, the team resolved the pathways of sulfur ionic migration, with grain boundaries in the nanocrystalline ReSe₂ layer confining sulfur ionic transport at the atomic scale1

1

. This enables deterministic control over synaptic weight updates, similar to the gating mechanism of biological ion channels. The device demonstrated key synaptic functionalities including multi-level conductance modulation, long-term potentiation/depression, paired-pulse facilitation, and a tunable short-term to long-term memory transition2

2

. The nanocrystalline ReSe₂ device exhibited a 34.7% increase in retention efficiency during learning-forgetting-relearning cycles compared to bulk ReSe₂1

1

.

Real-World Applications in AI Hardware

In system-level evaluations, the device successfully performed edge detection on natural images and achieved 96.24% classification accuracy on the CIFAR-10 image recognition task1

1

. This performance demonstrates the device's potential as a materials platform for AI hardware that processes visual information. "This study demonstrates a single-step method to design the structure of van der Waals crystals for optoelectronic synaptic devices that learn and store information using light," said Professor Taesung Kim2

2

. By structurally resolving the random nature of ionic migration and interfacial issues inherent in conventional devices, this architecture can be applied to research on next-generation neuromorphic semiconductors.

Implications for Neuromorphic Vision Systems

Rapid advancements in artificial intelligence and hyper-connectivity require neuromorphic vision systems capable of sensing and processing vast amounts of visual data in real time1

1

. Optoelectronic synapses, which exhibit conductance variations in response to light signals, serve as core components of these systems. The research, published in Advanced Materials with an impact factor of 26.8, was conducted as a collaborative effort among researchers from Sungkyunkwan University, the Center for Quantum Nanoscience at the Institute for Basic Science, and the Korea Institute of Machinery and Materials1

1

. The study received financial support from the National Research Foundation of Korea Leader Research Program, the Institute for Basic Science, and the Semiconductor-Track Graduate School Program funded by the Ministry of Trade, Industry and Energy2

2

.